Big ideas, open source innovation, and real-world impact.
This month, we brought the maker community together in Dallas and Mountain View to explore the future of Agentic AI and model validation.
From LLM risks in finance (led by Agus Sudjianto) to autonomous enterprise
RESEARCH
-
H2O Brings Maker Community Together for Agentic AI Innovation
By
–
-
Understanding Basic Artificial Intelligence Terminology and Concepts
By
–
What does artificial intelligence really mean? Understanding some basic
terminology will help you get your bearings: https://
okt.to/YngdWN @mvollmer1 @nicochan33 @blueyonder -
PyTorch Moves IO to TorchCodec for Better Efficiency
By
–
the IO part is moving to torchcodec: https://
github.com/pytorch/torchc
odec
…
We're not dropping IO without a replacement.
torchcodec is even more efficient than the torchaudio IO -
Parallel Approaches and Play Styles in AI Development
By
–
Next up, we need many parallel approaches and using different harnesses. Would be fun to see how different play styles manifest over time!
-
Stanford AI Lab Postdoctoral Fellowship Program Application Deadline
By
–
Tag someone who should apply for the Stanford AI Lab Postdoctoral Fellows Program! The deadline for applications is approaching soon
-

Meta’s Token-Shuffle Advances Autoregressive Image Generation
By
–
Even if Yann LeCun isn’t a fan of autoregressive models, Meta’s still pushing out some seriously impressive research using them.
Token-Shuffle: Towards High-Resolution Image Generation with Autoregressive Models
Paper: https://
arxiv.org/pdf/2504.17789 -

PhD Completion and New Role at Sakana AI Labs
By
–
Belated Life Update: Finished my PhD & joined @SakanaAILabs recently! I've been working on AI Scientist-v2 and will be giving two invited talks on Sat and Mon at #ICLR2025! Looking forward to discussing AI-driven scientific discovery & LLM agents!
-

Hassabis vs LeCun: Diverging Views on AGI Timeline
By
–
Nous avons maintenant une vision plus claire de la façon dont deux des plus gros laboratoires d’IA perçoivent l’AGI en interne : – Demis (Hassabis) prédit que l’AGI arrivera dans 5 à 10 ans et qu’elle sera révolutionnaire. – Yann LeCun pense que l’AGI est encore très loin et
-

Developer Competition: Build AI Breakthroughs and Climb Leaderboard
By
–
There's still time to engage! Connect with other devs Climb the leaderboard Earn credits + serious bragging rights Then, go create the next AI breakthrough. We’re watching. Jump in
-

LLMs Prioritize Reasoning Patterns Over Data Accuracy
By
–
STRUCTURE > ACCURACY: Models fed 100% WRONG answers only lost 3.2% performance!
LLMs care more about reasoning patterns than correct answers. We're using this insight to help enterprises train better models with less-than-perfect data. Read more: https://
bit.ly/44rkxW0